You have a recorded conversation and need text from it that can be trusted. Below: exactly what you get, what you do not, and what happens to the file.
A recorded conversation from a phone, a dictaphone or a messenger becomes verbatim text in minutes. Every passage carries a speaker label and the time from the start of the recording; hesitations, repetitions and unfinished sentences stay in. A conversation up to nineteen minutes costs $1.99 as a one-off, with no account and no card kept on file. Once signed in, you can delete the audio right after transcription.
Upload a recording. Create an account only after you see the result.
Drop a file here or choose one from your device. We will show you the beginning of the transcript first.
Which transcription model gets the fewest words wrong?
skryba.ai transcribes on Scribe v2, which gets 2.2% of words wrong in Artificial Analysis' independent index — the lowest of every model measured. Gemini 3.5 Transcribe gets 2.6% wrong, Whisper large-v3 4.1%. It is the same model that writes the transcript preview you see before paying.
Word Error Rate — words transcribed incorrectly — lower is better
Model
WER
skryba.ai (Scribe v2)the model we transcribe on
2.2%
MAI-Transcribe-1.5Microsoft Azure
2.4%
Gemini 3.5 TranscribeGoogle
2.6%
Universal-3 ProAssemblyAI
3.1%
GPT TranscribeOpenAI
3.3%
Whisper large-v3OpenAI
4.1%
Nova-3Deepgram
5.2%
Source: Artificial Analysis, AA-WER v2 index, non-streaming mode — read 27 August 2026. Seven of the 40-plus models measured: the top of the table and the names you already know. The index is measured mostly on English audio; for other languages, Polish included, no comparable public ranking exists. That is why we show you the start of every transcript before we ask for anything.
What you get: a verbatim record, not a summary
The default transcript is verbatim. Every “erm”, repetition, unfinished sentence and swear word stays in. Nothing is smoothed, reordered or filled in — if someone started a sentence three times, the text has three beginnings.
A split by speaker — the model tells the voices apart, and you name them in the editor: “me”, “him”, a first name.
The time from the start of the recording on every passage, in minutes and seconds.
An editor with the player beside it — click a sentence, hear that part of the audio.
Export to SRT (with the time of every passage) and TXT (with speaker labels or without).
A share link for a lawyer — it expires after 30 days and you can revoke it sooner.
What this record does not contain: the date and time of the conversation. The transcript only knows time counted from the start of the file, because the audio itself does not say when it was made. You document the date yourself — a screenshot of the recording list on the phone, the file's properties, the message the recording arrived in.
There is also a cleaned-up version, which removes fillers and stammers. For a recorded conversation you usually do not want it: a record meant to be checked against the audio should match it word for word.
What file this works from
Recorded conversations leave phones in a handful of formats, and none of them needs converting. Upload the file exactly as you have it.
Source
What you usually get
Note
Voice Memos on an iPhone
M4A
Share the file to a computer or upload straight from the phone
A dictaphone app or call recording on Android
M4A, AMR, 3GP, MP3
An Android call recording can be quiet on the other party's side — the preview shows whether it is enough
A WhatsApp voice message
OGG or OPUS
Export the message as a file, not as a screenshot
A digital dictaphone
MP3, WAV, WMA
Copy it over a cable and upload
A phone video
MP4, MOV
Upload it whole, no need to extract the audio
Scroll the table sideways to see the remaining columns.
The quality of the record depends on the quality of the recording more than on anything else. Two people in a room come out well; a phone in a pocket, an argument in several voices at once, or a call on a car speaker — worse, and then some words need fixing against the audio. The free preview shows the start of the record on your recording, so you see that before paying.
What it costs
You pay per started minute: $0.10 a minute, with a $1.99 minimum. That minimum covers conversations up to nineteen minutes, which is most recorded conversations. You pay once, by card or BLIK, with no account — after payment you read the whole thing at once, and the recording stays available for 30 days. Recordings up to five minutes are entirely free once you sign in.
Recording
One-off cost
Up to 5 minutes
$0 once signed in (15 free minutes a day)
6–19 minutes
$1.99
30 minutes
$3.00
An hour
$6.00
A dozen conversations from one matter
The 5-hour pack for $19.99, no expiry
Scroll the table sideways to see the remaining columns.
If there are more recordings — several conversations from the same matter, voice messages spanning months — an hour pack works out cheaper than paying for each one, and pack hours never expire. For 14 days after buying you can withdraw and get the whole amount back, including for minutes already used.
A recorded conversation is usually the most private thing anyone uploads to the internet. Below are the facts, without our interpretation.
The file goes into a private bucket on Cloudflare R2 within the European Economic Area. It is not published or indexed anywhere.
Transcription is fully automatic. Speech recognition is done by ElevenLabs, Inc. in the USA: it receives the file through an expiring link and returns the text; the transfer rests on standard contractual clauses — details in the privacy policy.
Once signed in, you delete a recording yourself, at any time, together with its transcript and every share. You can also switch on auto-deletion of audio after 1, 7 or 30 days — the transcript stays.
A recording uploaded without an account disappears by itself after 30 days, together with its transcript — that is how long you have to come back for it, pay for it or attach it to an account. Once you sign in, it stays until you delete it, or until the setting above deletes it for you.
A share link works for 30 days and you can revoke it sooner. The recipient sees the transcript; they do not get the file.
The cleaned-up version sends the text alone to Anthropic in the USA, and we ask about that separately, before the first use. If you never run it, the text goes nowhere beyond the places listed above.
If the recording is headed for a lawyer or a court
The transcript does not replace the recording. The evidence is the audio file; the written record lets you read it, quote it and point at the minute something was said — instead of scrubbing. Courts in Poland admit recordings made by a participant in the conversation, but the assessment of a particular recording belongs to the court in the particular case, so whether and how to present it is a conversation to have with your lawyer, not with us.
Keep the original file unchanged. Upload a copy — the original stays with you, on the device it was made on.
The transcript is produced automatically, so listen through it against the audio and fix misheard words in the editor. The TXT file carries a note that the text was generated by AI.
Name the speakers the way you will name them in your filing. Do not guess — if you are not sure who is speaking in a passage, leave a generic label.
Attach the time from the record (minute:second from the start of the file) to your filing, and describe the date and circumstances of the recording separately.
Is the transcript verbatim, or do you tidy up what was said?
It is verbatim. Hesitations, repetitions, unfinished sentences and swearing stay in, and nothing is reordered or filled in. The only corrections are the ones you make yourself in the editor, say when the model misheard a surname. The cleaned-up version, which removes fillers, exists separately and is usually not what you want for a recorded conversation.
Will the transcript show the date and time of the conversation?
No. The transcript carries time counted from the start of the recording — the minute and second of every passage — because the audio itself does not say when it was made. You document the date and time yourself: a screenshot of the recording list on the phone, the file's properties, or the message the recording came in.
Do you accept a WhatsApp voice message or an Android recording?
Yes, with no conversion. WhatsApp exports voice messages as OGG or OPUS, Android records in M4A, AMR or 3GP, iPhone in M4A, and dictaphones in MP3, WAV or WMA. Upload the file exactly as you have it. Android phone-call recordings can be quiet on the other party's side — the free preview shows whether the quality is enough.
Does anyone at your end listen to my recording?
Not to produce the transcript: speech recognition is done by a machine, and nobody listens to the recording in order to write it out. The file sits in a private bucket within the European Economic Area and is not published anywhere; the extent and purposes of the operator's access to data are set out in the privacy policy. Without an account, a recording disappears by itself after 30 days; once signed in you delete it at any time, including right after downloading the text — the transcript then stays on your account until you delete it too.
Do I need an account to buy the transcript of one conversation?
No. You upload the file, read the start of the transcript, pay by card or BLIK and read the whole thing — with no account to create and no card kept on file; your email goes only into the payment, for the receipt. A recording stays available for 30 days in the same browser, paid or not; if you sign in within that time, it is attached to your account and stops expiring (you can then delete the audio yourself, or set it to delete automatically). An account is only needed for free recordings up to five minutes.
Is a transcript like this enough as evidence?
The transcript alone is not — the evidence is the recording, and the written record helps read and quote it. In Poland, courts admit recordings made by a participant in the conversation, but each one is assessed by the court in the particular case. Keep the original file, fix misheard words against the audio, and discuss how to present the recording with your lawyer. The details are in a separate article.
